4 ms·

(Author chiming in, you've given me an opportunity for one of my favorite sport-arguments) I'll admit that even Jean-Louis Gassée (Former Mac head, Be founder)

(Author chiming in, you've given me an opportunity for one of my favorite sport-arguments) I'll admit that even Jean-Louis Gassée (Former Mac head, Be founder) later said NeXT+Steve was the right choice (see: https://9to5mac.com/2011/11/11/gassee-thank-god-apple-chose-steve-jobss-next-over-my-beos/ https://9to5mac.com/2011/11/11/gassee-thank-god-apple-chose-... ) - though he said it for personal reasons based on his contempt for Apple's management. Some not-carefully-composed points of argument that Be might have been the right choice, because I like doing it for sport: BeOS was truly a ground-up creation, OpenStep was a Display PostScript environment largely licensed from Adobe running on a hacked-up Mach with some BSD pieces pulled inside the kernel boundary to make performance acceptable (this is XNU, the kernel Apple still uses in everything). That had two points for Be - Apple had to replace the whole display layer, partly for licensing cost reasons and partly because developers weren't interested in dealing with the "Yellow Box" OpenStep API or being stuck in the "Blue Box" legacy compatibility jail. Equally or more important, BeOS was vastly more performant on the same hardware - doing the "the UI stays responsive while I'm playing a dozen media files back on this shitty Pentum MMX" stunt was always a blast around the turn of the century. BeOS was also naively SMP with a nifty and mostly pleasant thread model that has some advantages (and annoyances) over the dominant event loop way of things, but always remained responsive, while Apple inherited the spinning beach ball from NeXT. For another big technical point that history actually agrees with me on, OpenStep used a passable UFS implementation (that Apple reverted to HFS+, which was already getting long in the tooth in the late 90s, apparently largely because Adobe wouldn't get their shit together to make their products run on a case-sensitive FS...), but BeFS was lovely in ways we're finally getting elsewhere 20+ years later - exactly so in fact, Apple eventually hired Dominic Giampaolo who designed BeFS to design APFS, and it really shows. BeOS was also already suited for semi-embedded applications, both in various commercial multimedia gadgets, and also the BeIA internet appliance stuff. Since the bulk of modern NextStep-family devices are phones and tablets, there's no reason to think a Gasse/Be Apple wouldn't have thrived in this environment (I know, Palm borrowed Be parts for Palm OS Cobalt and didn't do well, but maybe they were already doomed by that time). Also, Be was _profoundly_ personal, looking toward a "everyone has their own WebServer running their personal services" kind of future, rather than the "Everyone back into the ~~mainframe~~ major cloud provider" future we got, and I can't help but thinking having one of the big industry players not in the roach-motel platform game would have been good for society. Jobs did in fact use a ThinkPad running OpenStep/Rhapsody builds around the time of the NeXT-buys-Apple-with-Apple's-Money situation. It was a 560 (not E), there is a NextComputers forum thread about it where Marc J. Driftmeyer who set up the machine Steve had at Macworld '97 talks about it http://www.nextcomputers.org/forums/index.php?topic=457.0 http://www.nextcomputers.org/forums/index.php?topic=457.0 .

DonHopkins 6y ago
This is the first time I've heard of the "NeXT-buys-Apple-with-Apple's-Money" meme. That's hilarious! What's the back story behind that? Reminds me of the old joke I heard years before I went to work at Kaleida, which then became my day-to-day reality: Q: What do you get when you cross Apple and IBM? A: IBM.
- PAPPPmAc 6y ago
  The short version is most of Apple's management was replaced with NeXT folks within a few years, and the combined company behaved more like NeXT than Apple. The degree to which Apple in the 90s was dysfunctional is pretty legendary, they started trying to replace the creaking 68k, cooperative multitasking hacked in after the fact classic Mac OS before 1988 and were still shipping what were supposed to be short-term life support releases until 2001 (...and really until 2002 because 10.0 on available hardware was miserable). They kept losing their replacement plans to development hell. A/UX was a nifty Mac/UNIX hybrid that came out in '88, and it was going to be the future, but was expensive (real UNIX license), deeply tied to 68k hardware, and the 4.0 PowerPC port never came out because it collapsed under the weight of "Build a combined OSF platform out of A/UX and (IBM's) AIX that takes over from both and rules the whole computing market" ambitions. That joint project also included Taligent, an OOP-platform-thing that is basically the benchmark for failed software development. There was also Copland, which was a separate in-house microkernel-based OS that was supposed to be MacOS 8, developed between '94-96... that the Platinum theme was the only survivor from. I think part of the reason they _did_ go with NeXT is OpenStep looked a lot like the Pink/Taligent platform plan, but not bungled. ...I have a draft of a few-page humorous telling of the "How Apple and IBM spent 15 years failing to write OSes" story but I keep tweaking it and need to dig up my copy of _Apple: The Inside Story of Intrigue, Egomania, and Business Blunders_ to add a few references, and find some better reference material for the OS/2 fiasco between IBM and Microsoft that is one of the core "all this 90s tech industry dysfunction fits together" points most of the tellings miss.
spijdar 6y ago
I'll bite, since I wasn't around at the time and am not deeply attached to either, but have some thoughts none-the-less for sport. I'm going to "flip" several of your points and "reword" them from the perspective of a dirty NeXT sympathizer ;) OpenStep was built on top of a then rather battle hardened Objective-C toolchain and "IDE" which had been in development since the 80s, and was an open standard that would target both the Mach OS, Solaris, and Windows NT. Mach 2 always shipped with a full BSD kernel running as a server -- the modifications were such that BSD would request resources like memory and process time from the mach kernel. This is why for many years Mach based systems kicked the butts of other BSDs. While CSRG's BSD kept its awful virtual memory system from VAX that crapped out after 30-40 MB of address space, Mach's VM system was scalable and performant. (Mach 3 and above were "more proper", and indeed Darwin's combination of OSF/1's Mach derivative and 4.4BSD-lite's kernel/userland was for performance reasons.) Mach also gave NeXT a threading model, something absent from BSD. It is exploited by the objective-C framework and the driver layer natively. https://src.paperfox.es/next.jpg https://src.paperfox.es/next.jpg Because of its BSD roots, NeXT shipped with support for multiple users (something Haiku still lacks real support for to this day, with everything "Be" related stuck under a single user, although the underlying "not-unix-but-POSIX" layer does support it), and a mature BSD sockets library, along with its own (horrific, but there!) interpretation of domain services, NetInfo. Be shipped with a pseudo-unix environment, with a fair amount of POSIX represented, and a bash shell and GNU utilities preinstalled, albeit lacking any concept of users or permissions... And to me that's the kicker -- both OSes intentionally had "UNIX layers", Be's was just home-grown and lacked features compared to OpenStep's "4.4BSD but with patches and significantly enhanced resource handling" environment, with much greater software compatibility as a result. NeXT had the advantage of being an older piece of software, with a much greater depth of "backwards compatibility" while also having a fairly well designed native framework with a well-represented stack of native software. If I was given some (ridiculous) mission of building <x> desktop OS from either BeOS R5's kernel/low level userland or some late release of Open/Rhapsody/Darwin, it'd be a no-brainer to me. (as a final thought, it's curious you'd mention running a personal web server as an example, when the first web server was developed on NeXT) (I also ignored Be's filesystem because it is, frankly, superior, however I believe it's not enough by itself to change the merit of the entire OS, since its value either depends on developer buy-in or how much utility it gives to the OS shell, and OS X implemented similar features within Finder using indexers and databases. Ugly, but if it works...)

DonHopkins 6y ago

This is the first time I've heard of the "NeXT-buys-Apple-with-Apple's-Money" meme. That's hilarious! What's the back story behind that? Reminds me of the old joke I heard years before I went to work at Kaleida, which then became my day-to-day reality: Q: What do you get when you cross Apple and IBM? A: IBM.

PAPPPmAc 6y ago
The short version is most of Apple's management was replaced with NeXT folks within a few years, and the combined company behaved more like NeXT than Apple. The degree to which Apple in the 90s was dysfunctional is pretty legendary, they started trying to replace the creaking 68k, cooperative multitasking hacked in after the fact classic Mac OS before 1988 and were still shipping what were supposed to be short-term life support releases until 2001 (...and really until 2002 because 10.0 on available hardware was miserable). They kept losing their replacement plans to development hell. A/UX was a nifty Mac/UNIX hybrid that came out in '88, and it was going to be the future, but was expensive (real UNIX license), deeply tied to 68k hardware, and the 4.0 PowerPC port never came out because it collapsed under the weight of "Build a combined OSF platform out of A/UX and (IBM's) AIX that takes over from both and rules the whole computing market" ambitions. That joint project also included Taligent, an OOP-platform-thing that is basically the benchmark for failed software development. There was also Copland, which was a separate in-house microkernel-based OS that was supposed to be MacOS 8, developed between '94-96... that the Platinum theme was the only survivor from. I think part of the reason they _did_ go with NeXT is OpenStep looked a lot like the Pink/Taligent platform plan, but not bungled. ...I have a draft of a few-page humorous telling of the "How Apple and IBM spent 15 years failing to write OSes" story but I keep tweaking it and need to dig up my copy of _Apple: The Inside Story of Intrigue, Egomania, and Business Blunders_ to add a few references, and find some better reference material for the OS/2 fiasco between IBM and Microsoft that is one of the core "all this 90s tech industry dysfunction fits together" points most of the tellings miss.

spijdar 6y ago

I'll bite, since I wasn't around at the time and am not deeply attached to either, but have some thoughts none-the-less for sport. I'm going to "flip" several of your points and "reword" them from the perspective of a dirty NeXT sympathizer ;) OpenStep was built on top of a then rather battle hardened Objective-C toolchain and "IDE" which had been in development since the 80s, and was an open standard that would target both the Mach OS, Solaris, and Windows NT. Mach 2 always shipped with a full BSD kernel running as a server -- the modifications were such that BSD would request resources like memory and process time from the mach kernel. This is why for many years Mach based systems kicked the butts of other BSDs. While CSRG's BSD kept its awful virtual memory system from VAX that crapped out after 30-40 MB of address space, Mach's VM system was scalable and performant. (Mach 3 and above were "more proper", and indeed Darwin's combination of OSF/1's Mach derivative and 4.4BSD-lite's kernel/userland was for performance reasons.) Mach also gave NeXT a threading model, something absent from BSD. It is exploited by the objective-C framework and the driver layer natively. https://src.paperfox.es/next.jpg https://src.paperfox.es/next.jpg Because of its BSD roots, NeXT shipped with support for multiple users (something Haiku still lacks real support for to this day, with everything "Be" related stuck under a single user, although the underlying "not-unix-but-POSIX" layer does support it), and a mature BSD sockets library, along with its own (horrific, but there!) interpretation of domain services, NetInfo. Be shipped with a pseudo-unix environment, with a fair amount of POSIX represented, and a bash shell and GNU utilities preinstalled, albeit lacking any concept of users or permissions... And to me that's the kicker -- both OSes intentionally had "UNIX layers", Be's was just home-grown and lacked features compared to OpenStep's "4.4BSD but with patches and significantly enhanced resource handling" environment, with much greater software compatibility as a result. NeXT had the advantage of being an older piece of software, with a much greater depth of "backwards compatibility" while also having a fairly well designed native framework with a well-represented stack of native software. If I was given some (ridiculous) mission of building <x> desktop OS from either BeOS R5's kernel/low level userland or some late release of Open/Rhapsody/Darwin, it'd be a no-brainer to me. (as a final thought, it's curious you'd mention running a personal web server as an example, when the first web server was developed on NeXT) (I also ignored Be's filesystem because it is, frankly, superior, however I believe it's not enough by itself to change the merit of the entire OS, since its value either depends on developer buy-in or how much utility it gives to the OS shell, and OS X implemented similar features within Finder using indexers and databases. Ugly, but if it works...)